Improved total atmospheric water vapour amount determination from near-infrared filter measurements with sun photometers

In this work we explore the effect of the contribution of the solar spectrum to the recorded signal in wavelengths outside the typical 940-nm filter's bandwidth. We use gaussian-shaped filters as well as actual filter transmission curves to study the implications imposed by the non-zero out-of-band contribution to the coefficients used to derive precipitable water from the measured water vapour band transmittance. The moderate-resolution SMARTS radiative transfer code is used to predict the incident spectrum outside the filter bandpass for different atmospheres, solar geometries and aerosol optical depths. The high-resolution LBLRTM radiative transfer code is used to calculate the water vapour transmittance in the 940 nm band. The absolute level of the out-of-band transmittance has been chosen to range from 10 super(− 6) to 10 super(− 4), and typical response curves of commercially available silicon photodiodes are included into the calculations. It is shown that if the out-of-band transmittance effect is neglected, as is generally the case, then the derived columnar water vapour is systematically underestimated by a few percents. The actual error depends on the specific out-of-band transmittance, optical air mass of observation and water vapour amount. We apply published parameterized transmittance functions to determine the filter coefficients. We also introduce an improved, three-parameter, fitting function that can describe the theoretical data accurately, with significantly less residual effects than with the existing functions. Further investigations will use experimental data from field campaigns to validate these findings.